Artificial intelligence is quickly becoming central to both security and economic strength. The United States must act decisively to lead this transition by ensuring that U.S. technology undergirds global AI use and that adversaries cannot easily abuse advanced AI. In the wrong hands, powerful AI systems have the potential to exacerbate significant national security risks, including by enabling the development of weapons of mass destruction, supporting powerful offensive cyber operations, and aiding human rights abuses, such as mass surveillance. Today, countries of concern actively employ AI - including U.S.-made AI - in this way, and seek to undermine U.S. AI leadership.
To enhance U.S. national security and economic strength, it is essential that we do not offshore this critical technology and that the world's AI runs on American rails. It is important to work with AI companies and foreign governments to put in place critical security and trust standards as they build out their AI ecosystems.
To strengthen U.S. security and economic strength, the Biden-Harris Administration today is releasing an Interim Final Rule on Artificial Intelligence Diffusion. It streamlines licensing hurdles for both large and small chip orders, bolsters U.S. AI leadership, and provides clarity to allied and partner nations about how they can benefit from AI. It builds on previous chip controls by thwarting smuggling, closing other loopholes, and raising AI security standards.
Six key mechanisms in the rule catalyze the responsible diffusion of U.S. technology:
- No restrictions apply to chip sales to 18 key allies and partners. This flexibility enables jurisdictions with robust technology protection regimes and technology ecosystems aligned with the national security and foreign policy interests of the United States to benefit from seamless large-scale purchases.
- Chip orders with collective computation power up to roughly 1,700 advanced GPUs do not require a license and do not count against national chip caps. The overwhelming majority of chip orders are in this category, especially those being placed by universities, medical institutions, and research organizations for clearly innocuous purposes. Streamlined processing of these orders represents an improvement over the status quo, rapidly accelerating low-risk shipments of U.S. technology around the world.
- Entities that meet high security and trust standards and are headquartered in close allies and partners can obtain highly trusted "Universal Verified End User" (UVEU) status. With this status, they can then place up to 7% of their global AI computational capacity in countries around the world - likely amounting to hundreds of thousands of chips. This trusted status is granted on a global and enduring basis, allowing responsible entities to expand rapidly and flexibly, and strengthening U.S. and allied global leadership while keeping frontier training at home.
- Entities that meet the same security requirements and are headquartered in any destination that is not a country of concern can apply for "National Verified End User" status, enabling them to purchase computational power equivalent to up to 320,000 advanced GPUs over the next two years. This provision permits trusted national entities to benefit from advanced U.S. technology, serving local, governmental, and regional customers, while guarding against diversion risks.
- Non-VEU entities located outside of close allies can still purchase large amounts of computational power, up to the equivalent of 50,000 advanced GPUs per country. This cap ensures that U.S. technology is available to service foreign governments, healthcare providers, and other local businesses.
- Government-to-government arrangements cultivate an international ecosystem of shared values regarding the development, deployment, and use of AI. Governments that sign these arrangements - which align those nations' export control, clean energy, and technology security efforts with the United States - can double their chip caps (up to 100,000 of today's advanced GPUs).
Even as it encourages the diffusion of U.S. technology, the rule takes significant steps against countries of concern, constraining them from accessing advanced AI systems and the computing power used to train them. These actions include:
- Continuing to ensure that advanced semiconductors sold abroad are not used by countries of concern to train advanced AI systems, while still permitting access for general-purpose applications from telecommunications to banking.
- Restricting the transfer to non-trusted actors of the model weights for advanced closed-weight models. The rule does not in any way inhibit the publication of model weights for open-weight models.
- Setting security standards to protect the weights of advanced closed-weight AI models, permitting them to be stored and used securely around the world while helping prevent illicit adversary access.
The rule builds on previous regulations that aim to protect U.S. national security, including the October 2022 and October 2023 chip controls. It follows a broad range of relevant engagements over the past ten months with stakeholders, bipartisan members of Congress, industry representatives, and foreign allies and partners.
